ECG-based biometric under different psychological stress states.

Area of Science:

• Biometrics
• Signal Processing
• Psychophysiology

Background:

• Electrocardiogram (ECG) biometrics are explored for identification.
• ECG signals are sensitive to psychological stress, challenging traditional verification methods.
• Existing ECG biometrics struggle with varying psychological stress states.

Purpose of the Study:

• To propose a hybrid feature method for ECG biometrics resilient to psychological stress.
• To introduce a Stress Classification Coefficient (SCC) to quantify stress effects on heart rate variability (HRV).
• To enhance the robustness and applicability of ECG-based biometrics across different stress levels.

Main Methods:

• Manual feature extraction involves HRV analysis and Gaussian Mixture Model (GMM) for mental state assessment.
• A one-dimensional convolutional neural network (1D-CNN) is used for automatic feature extraction from ECG signals.
• Hybrid features from manual and automatic methods are fused and classified using a Support Vector Machine (SVM).

Main Results:

• The method was tested on 23 healthy students under induced stress conditions.
• Achieved an average recognition rate exceeding 95%.
• Attained an average F1 score of 0.97, demonstrating high accuracy.

Conclusions:

• The proposed hybrid feature approach effectively addresses the challenge of psychological stress in ECG biometrics.
• This method offers a viable solution for reliable ECG-based identification under diverse psychological stress conditions.
• The study paves the way for more robust and adaptable ECG biometric systems.